Re: [office-metadata] Groups - Metadata_Model_Proposal (07-02-19-ODF-MetaData.odt) uploaded

[Elias Torres <eliast@us.ibm.com>]

Michael.Brauer@Sun.COM wrote on 02/21/2007 08:34:41 AM:

> Hi,
>
> I won't be able to attend to the call today. Below are my comments:
>
> Bruce D'Arcus wrote:
> >
> > On Feb 20, 2007, at 10:57 AM, Svante Schubert wrote:
> >
> >> Hi Bruce,
> >>
> >>>
> >>> 1) Section 1.2.2
> >>>
> >>> a. The metadata attributes for about and property need to be in a
> >>> namespace other than the rdf namespace. I have typically been
> >>> assuming meta.
> >> Yes, you might be correct on this. Actually, we swifted the name in
> >> the last hour from 'meta:id' to 'rdf:about', as we wanted to reuse
> >> existing tags.
> >> I assume the rdf:about is used in a slightly different way in RDF.
> >> What we intended to do was to 'tag'/assign an IRI to the ODF element.
> >> Therefore it might be used multiple times from various vocabularies,
> >> which simply would relate to this ODF element via this IRI. As we
> >> described by this IRI always an ODF element, we suggested even to
> >> create a set of IRI for our purposes, to make this RDF subject/object
> >> group more distinguishable.
> >> Perhaps you might give us some more details about the problem as you
> >> see it?
> >
> > OK, there are two aspects to this issue: the easy part, and the hard
part.
> >
> > The easy part is just that for in content metadata we don't want to use

> > the rdf namespace. That's what I was referring to above.
>
> We can use the RDF namespace for an attribute in the content.xml if, and
> only if, an approved RDF specification specifies the attribute in
> question as a global attribute that can be used in other schemas. If
> this is not the case, we must define an attribute in a namespace that
> belongs us. We may however use an "rdf" namespace prefix in the
> specification for this namespace if we think that this would be
> appropriate. We do so for instance for the attributes we borrow from
> XSL-FO and SVG.

I think that using rdf:about per se is not incorrect. We could stretch it
that far and use it, but I think we need more than just about, we need
property, class, etc. Since we don't control that namespace, I'd think it'd
not be the best solution to have one attribute from rdf namespace and the
rest from our own. Hence we should put all of our attributes in our own
schema. If RDFa was published and at first we wanted it to be a vocabulary
for all XML dialects, then we could have used rdfa:about, etc, but
unfortunately that's not the case. I also think that it's not necessary
that we re-use the rdf ns prefix if it's not rdf. I think our spec is
office metadata so something along those lines might be more intuitive than
just rdf like meta, odf-meta, ometa, whatever.

>
> >
> > The hard part is the identifier stuff: what kind of a URI (local vs.
> > global) do we assign to content nodes, to represent what? Does it
> > represent the ODF document object (paragraph, table, etc.) or does it
> > represent the content of that object? When is it valuable to identify
> > these items?
> >
> > This is a deep philosophical question, ultimately, but we need to make
a
> > clear statement on it.
>
> It's clear to me that there is a difference, but do we really have to
> decide for the one or the other? Could we make this dependent from the
> predicates that are used, that is, assume that the predicates themselves
> specify whether they are about the ODF object or its content?

I don't think dependency on predicates give us a good solution. That would
require us to maintain a never ending list of predicates to know the
meaning of an identifier.  Even worse, what if a predicate applies to both?

Borrowing from my RDFa experience to answer some of Bruce's questions:

> > The hard part is the identifier stuff: what kind of a URI (local vs.
> > global) do we assign to content nodes, to represent what?

local vs global doesn't matter because in the end they must be absolute
meaning global. We do have options like for example grabbing xml:id="foo"
and creating some magical URI that references this document after
processing. In other words, we can come up with some shortcuts or
processing rules to make it more elegant (sort-of like what Floria is
doing, but we really need to understand the big picture before we come up
with the shortcuts so we don't leave any holes unplugged).

> > Does it
> > represent the ODF document object (paragraph, table, etc.) or does it
> > represent the content of that object?

In RDFa, xml:id identifies the element and so does <a name=""/> but
rdf:about is used to "establish" the subject at that element or its
children of a triples which uses the "content" of the element. meta:about
doesn't identify the element or the content. I think this is extremely
important for us to agree on.

>
> If we have to decide on one or the other option: If we say the URI
> represents the ODF object, then we still should be able to make
> statements about the content of that object by using predicates that are
> defined to be about the content. If we say the URI represents the
> content of the object, are we able to say anything about the object
itself?

Please read above.

>
>
> >
> >>> 3) On this comment:
> >>> [Patrick: We might want to create a own IRI ODF Schema]
> >>>
> >>> I doubt we need to. I suggest we might list some good ones to use
> >>> (LSIDs, etc.)
> >> The idea behind was that we are always giving IRIs for ODF elements to

> >> describe them.. To make this clear for everyone, it might be good to
> >> use an own ODF element specific.
> >
> > OK, goes back to the need to define thee identifier smore clearly. It
> > sound analogous to using xml:id, yes?
>
> That actually is not clear to me, too. My understanding is that one
> assigns an IRI to objects in the content (or to the content of object -
> that does not matter here), and that ones uses this IRI in the rdf:about
> attributes of RDF-XML files to make statements about these objects.
>
> Is that correct?

Very close. Let me try to exemplify (again from RDFa perspective):

In content:

<text:span xml:id="foo">Hi Michael</text:span>

In package:

<rdf:Description rdf:about="content.xml#foo">
  <dc:title>Welcome Text</dc:title>
</rdf:Description>

In this case there we have an xml:id which is used to generate a (relative)
URI <content.xml#foo>. This URI can then be used in the RDF/XML to make
statements about the **subject** not the object. The subject is #foo.

<content.xml#foo> dc:title "Welcome Text".

I would make this triple mean the "foo" element's title (or label) is some
welcome text. It doesn't say anything about Hi Michael especifically.


Now

In content:

<text:span meta:about="bar" meta:property="cal:start">August 23,
2006</text:span>

In package:

<rdf:Description rdf:about="content.xml#bar">
  <cal:end>August 26, 2006</dc:title>
</rdf:Description>

When we use meta:about we are simply using it to give a subject to our
triple, not the element, not the content of the element.:

<content.xml#bar> cal:start "August 23, 2006" .

Additionally, we can make other statements about the same subject in our
RDF/XML.

I hope this really helps.

>
> If so, isn't it the case the IRI is just a unique name and is not
> interpreted at all?

That's correct and what we mean by opaque.

>
>
> Best regards
>
> Michael
>
> Best regards
>
> Michael